Scientific Multimodal MCQA on SciFIBench
58AccuracySCICON
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SCICONBackbone=Qwen 3.5 9B2026.03 | 58 | 57.55 | |
| VCDBackbone=Qwen 3.5 9B2026.03 | 55.9 | 55.55 | |
| GreedyBackbone=Qwen 3.5 9B2026.03 | 55.1 | 54.67 | |
| SCICONBackbone=Phi-3.5-vision-instruct2026.03 | 54.9 | 55.02 | |
| VCDBackbone=Phi-3.5-vision-instruct2026.03 | 53.5 | 53.75 | |
| ICDBackbone=Qwen 3.5 9B2026.03 | 53 | 52.13 | |
| SCICONBackbone=Qwen 3.5 4B2026.03 | 48.7 | 46.99 | |
| GreedyBackbone=Phi-3.5-vision-instruct2026.03 | 48.6 | 48.47 | |
| ICDBackbone=Phi-3.5-vision-instruct2026.03 | 47.1 | 46.86 | |
| GreedyBackbone=Qwen 3.5 4B2026.03 | 46.2 | 44.93 | |
| VCDBackbone=Qwen 3.5 4B2026.03 | 45.4 | 44 | |
| ICDBackbone=Qwen 3.5 4B2026.03 | 40.5 | 38.23 |